Professional Documents
Culture Documents
1
The cooperative optimal dispatch of cold loads and the voltage at different nodes, a model of the LV network
community energy storage unit has shown the effectiveness of supplying a community of 100 dwellings has been created.
cooperative resource dispatch, towards meeting the common For this work the CPFT is set at a constant level each day
objective function of maintaining community's power flow which is the average net consumption predicted for the
below a pre-specified threshold. community based on average consumption and average
renewable energy source (RES) production each day. For this
work, a year’s work of community loading has been extracted
from the electricity demand model at [6], and PV data is
available from local measurements with 5 min resolution,
providing community equipment usage and power flow data
which can be optimised by the CPFT. The structured
management of the resources is able to control the CPF close
to its CPFT, and in particular significantly reduces the
evening peaks [5].
2
comparison with the transmission network, the low voltage calculate the line currents, until the second bus is reached. At
distribution networks are mostly resistive [11], [12]. this point the last calculated line current is compared with its
previous value. If the difference between the two line current
values is less than the convergence tolerance (10-6) the loop
terminates, otherwise the backward sweep is executed. The
backward sweep starts at bus 2, where the value of the line
current between bus 2 and bus 3 calculated in the forward
sweep is used to determine a new value for voltage at bus 3.
Then the load current is simply calculated by applying Ohm’s
Fig. 2 Schematic of the candidate network, connecting a secondary law and the application of KCL at bus 3 gives the line current
transformer to 100 dwellings between bus 3 and bus 4. The backward sweep continues until
the last bus voltage (i.e. V100) is calculated. Upon termination
Therefore the feeder inductance is assumed to be zero. The of the backward sweep, the forward sweep is initiated using
impedance of the LV feeder depends on its cable the most recently calculated value of V100 as an input. The
characteristics (cross section, material), in addition to its execution of both the forward and the backward sweeps
overall length. Since the feeder length would determine its continues until the voltage variation at all the buses converges
overall resistance and the resultant voltage drop across the to a value less than the specified tolerance.
feeder, it is important to determine this parameter with respect
to the maximum tolerated voltage drop across the feeder After performing the Ladder Iterative load flow analysis
length. Therefore the maximum feeder length is calculated so technique on the network topology shown in Fig. 2, the
that the voltage drop at the feeder end node does not fall voltage magnitude at every bus across the feeder is calculated.
below the lower statutory limit of 216.2Vrms. This Fig. 3 shows the voltage variation obtained at the100 different
calculation includes the After Diversity Maximum Demand buses, with an approximate drop of 3% at bus 100.
(ADMD) which represents the average power per dwelling,
and the resistivity of the feeder conductor. A three phase
185mm2, mains LV cable with an aluminium core has been
selected from [13]. The value of ADMD obtained from [14] is
2kW and the maximum feeder length is calculated as 3.07km.
This feeder length has a total feeder resistance of 0.47Ω.
Assuming that the dwellings are equally distanced from one
another, every feeder section between different dwellings has
been assumed to have a resistance of 0.0047Ω. Having
calculated the total feeder length and impedance for the
network under investigation, Kersting’s ladder iterative load
flow analysis technique has been used to determine the Fig. 3. Voltage variation at different buses across the feeder with 1kW
voltage variation at every bus across the feeder with a time load in every dwelling
resolution of 1 second.
An entire day’s worth of simulation using the Ladder Iterative
The ladder iterative technique is an algorithm developed technique takes 18 minutes 43 seconds which is an acceptable
specifically to compute power flow for radial networks. This simulation speed for this work.
method relies heavily on application of Kirchhoff’s Current
Law (KCL) and Kirchhoff’s Voltage Law (KVL) at all 4 Simulation Scenarios
connection buses across the network, in a series of forward
and backward sweeps in order to converge to a solution [15]. As highlighted before, the electrification of heat and transport
This method is essentially comprised of two iterative in addition to the integration of distributed generation, are
calculation stages termed the forward and the backward envisaged to occur in the near future. However the integration
sweeps in which the line currents and bus voltages are of these emerging load and generation units are expected to
respectively determined. For the network topology under affect the operation of the distribution network, and would not
consideration shown in Fig.2, the forward sweeps starts from be possible unless the required degree of active management
the feeder end point (i.e. bus 100). In order to start the first and control is incorporated at different levels of the system. In
iteration the voltage is assumed to be equal to the reference particular significant voltage variation beyond the statutory
voltage of 1p.u. The value of the load power is also known at limits is expected for the radial low voltage distribution
every bus. It is then possible to calculate the line current in networks which could consequently lead to voltage
each bus. instability.
Having calculated the bus currents and assuming a value of 4.1 The Load Model
1p.u. for the voltage at bus 100, the voltage at the adjacent
bus can be calculated by applying KVL. These steps are In order to investigate the impact of different penetration
consecutively applied between adjacent nodes in order to levels of EV and DG on the low voltage distribution network,
3
and also examine the effectiveness of the CPFC algorithm at node with 30% EV, 50% GSHP and PV penetration levels. It
indirectly regulating the voltage at different nodes, a model of is worth noting that in order to investigate the worst case
the LV network supplying a community of 100 dwellings has scenario, the allocation of these resources starts from the
been created as described in [5]. The model runs a real time dwelling at the feeder end node.
simulation of 100 different dwellings. The electricity demand
model presented in [6] is used to quantify the variation of
electrical appliance load throughout the day. Bespoke time
domain models for each of these pieces of equipment are used
to obtain a year’s worth of loading data for each simulated
house, sampled at 1 minute intervals. The simulation also
includes LI-Ion battery, Photovoltaic Panels, Ground Source
Heat Pumps and Electric Vehicle models. The CPFC
algorithm can interact with equipment to shed or enforce
connect suitable devices based on the requirement to follow
the community power flow target. The community power
flow is quantified by adding the power from different DERs
(i.e. battery, PV, GSHP and EV) and the appliance load from Fig. 4. Voltage variation at three nodes across the feeder for a
community of 100 dwellings with no EV or DG
all households.
According to Fig. 5. when there is no control of active power
The electrical loads and DERs from each of the 100 house- (red trace), the voltage magnitude falls below the lower
models is connected to one of the nodes of the distribution statutory limit repeatedly (shown by the green dotted line of
system shown in Fig 2 to demonstrate the variation of voltage Fig 5.). This is caused as a result of EV charging during the
along the radial feeder with time during the day. evening hours. The voltage can reach a value as low as 200V,
if there is no control of active power.
4.2 Voltage Variation with No EV or PV
The application of the CPFC algorithm significantly reduces
The voltage drop across the LV feeder is directly proportional the occurrence of voltage excursions below the lower
to the feeder length and the distance from the secondary statutory limit (blue trace). However minor voltage
transformer. Therefore while the voltage at the connection excursions still occur during the interval 23:00 - 04:00 hours
node between the secondary transformer and the LV feeder of the second day. This is primarily due to simultaneous
could be effectively regulated close to the nominal value, the reconnection of electric vehicles for charging – all are
voltage at the feeder end node would suffer the most configured for overnight charging.
significant variation. This effect will be more profound when
there is either an increase in load or distributed generation, as
both cases entail an increase in feeder current and
consequently a significant voltage drop across the feeder.
4
when the EV penetration is increased to 90% as shown on
Fig.7. According to Fig.7 the regulated voltage with the
CPFC and 90% EV penetration drops as low as 201 V, which
might also cause instability.
5
generation from PV units is not as severe and only minor
voltage excursion above the upper voltage statutory limits are
observed even at 90% PV penetration. However it is
important to note that the impact of DG on voltage might be
more significant if DG units with higher rated power were
considered. In conclusion, the inclusion of voltage as a
secondary control parameter in the CPFC’s decision making
could ensure effective control of power and voltage
throughout the day. This could also avoid unnecessary DG
curtailment and has been suggested as future work from this
project.
Variation of voltage at the end of feeder at different PV penetration
levels
Acknowledgements
This work was supported by European Regional Development
Fund (ERDF), within the Accelerating Low Carbon Economy
(ALCE) project.
References
[1] P. Kundur, Power System Stability and Control, 1994.
[2] F. Castellanos and A. F. Mohammed, "Voltage stability and reactive
power compensation on the T&TEC system," in SoutheastCon, 2003.
Proceedings. IEEE, 2003, pp. 134-139.
Fig 9. Voltage variation at the feeder end node at 90% PV penetration [3] I. Grau, L. M. Cipcigan, N. Jenkins, and P. Papadopoulos, "Microgrid
intentional islanding for network emergencies," in Universities Power
Engineering Conference (UPEC), 2009 Proceedings of the 44th
It is evident that application of the CPFC algorithm has International, 2009, pp. 1-5.
indirectly resulted in effective voltage regulation very close to [4] Fazeli, A.; Christopher, E.; Johnson, C. M.; Gillott, M.; Sumner, M.,
the nominal value of 230 volts throughout the day. The CPFC "Investigating the effects of dynamic demand side management within
intelligent Smart Energy communities of future decentralized power
algorithm has effectively curtailed the correct magnitude of system," Innovative Smart Grid Technologies (ISGT Europe), 2011
PV generation, in order to control the CPF at its target value. 2nd IEEE PES International Conference and Exhibition on , vol., no.,
This real time control of active power has indirectly regulated pp.1,8, 5-7 Dec. 2011
the voltage at different nodes across the feeder. [5] A. Fazeli , Sumner, M, Johnson, C M, Christopher, E,, "Coordinated
Optimal Dispatch of Distributed Energy Resources within a Smart
Energy Com-munity Cell," presented at the 2012 3rd IEEE PES
It is however important to note that the DG curtailment Innovative Smart Grid Technologies Europe Berlin, 2012.
considered in this study would not be beneficial for the [6] I. Richardson, M. Thomson, D. Infield, and C. Clifford, "Domestic
consumer. On the other hand if a bulk battery with sufficient electricity use: A high-resolution energy demand model," Energy and
Buildings, vol. 42, pp. 1878-1887, 2010.
capacity were available within this community, the locally [7] P. J. Lagace, M. H. Vuong, and I. Kamwa, "Improving power flow
generated energy would be stored in the battery and not convergence by Newton Raphson with a Levenberg-Marquardt
curtailed. In that case both the consumer and the DNO would method," in Power and Energy Society General Meeting - Conversion
benefit from the operation of the CPFC algorithm and the and Delivery of Electrical Energy in the 21st Century, 2008 IEEE,
2008, pp. 1-6.
battery. That justifies the need for energy storage at [8] I. Richardson, "Integrated high-resolution modelling of domestic
community level. electricity demand and low voltage electricity distribution networks,"
PhD, Electrical and electronic engineering, Loughborough, 2011.
[9] Y. Kumar, K. Dwivedi, and G. Agnihotri, "Development of Ant
5 Conclusion Algorithm for load flow analysis," in Power Systems Conference and
Exposition, 2009. PSCE '09. IEEE/PES, 2009, pp. 1-5.
The impact of load and DG growth on voltage variation [10] D. Bhujel, B. Adhikary, and A. K. Mishra, "A Load Flow Algorithm
within a candidate section of the LV network supplying a for Radial Distribution System with Distributed Generation," in
community of 100 dwellings has been investigated. The Sustainable Energy Technologies (ICSET), 2012 IEEE Third
International Conference on, 2012, pp. 375-380.
simulation results show that, when there is no control of [11] T. L. Vandoorn, J. D. M. De Kooning, B. Meersman, and L.
active power, even at 30% EV penetration the voltage Vandevelde, "Soft curtailment for voltage limiting in low-voltage
magnitude falls below the lower statutory limit repeatedly and networks through reactive or active power droops," in Energy
can reach a value as low as 200V. On the other hand the Conference and Exhibition (ENERGYCON), 2012 IEEE International,
2012, pp. 545-550.
application of the CPFC algorithm significantly reduces the [12] Y. Vencken, W. Labeeuw, and G. Deconinck, "Dealing with an
occurrence of voltage excursions below the lower statutory overdose of photovoltaics at distribution level," in PowerTech
limit. However since the CPFC algorithm does not use (POWERTECH), 2013 IEEE Grenoble, 2013, pp. 1-5.
voltage as a control parameter it fails to regulate the voltage [13] eon, "Cables, Cable Laying & Accessories Manual " 2006.
[14] e. o. c. networks, "Network Design Manual."
at 60% and 90% EV penetration levels, as more significant [15] G. T. H. Alejandro D. Domínguez-García, Siddharth Suryanarayanan,
voltage excursions occur. This could lead to instability, and "Implications of the Smart Grid Initiative on Distribution
consequently voltage collapse. The impact of power Engineering," University of Illinois at Urbana/Champaign2011.